Leveraging Missing Data and Measurement Error Approaches in Propensity Score-Based Analyses of Real-World Data

Real-world data feature complex patterns of missing data arising due to the heterogeneous process of patients interacting with the healthcare system. This creates challenges for propensity score-based comparative effectiveness analyses since many potential confounders will be inconsistently available across patients. This talk presented an evaluation of measurement error and missing data approaches to this problem, motivated by a real-world study of the comparative effectiveness of treatments for advanced urothelial cancer, and concluded with reflections on how to embrace modern advances in data sources and methodology while preserving time-tested principles that support research rigor and reproducibility.